Running head: MISLEADING HEADLINES
****IN PRESS AT THE JOURNAL OF EXPERIMENTAL PSYCHOLOGY: APPLIED****
The Effects of Subtle Misinformation in News Headlines
Ullrich K. H. Ecker1, Stephan Lewandowsky1,2, Ee Pin Chang1, and Rekha Pillai1
1
2
School of Psychology, University of Western Australia
School of Experimental Psychology, University of Bristol
Word count: 8,783 (excluding title page, abstract, footnotes, references, figure captions, and
appendices); 13,870 (including footnotes, references, figure captions, and appendices)
Author Note:
This research was facilitated by a Discovery Grant and an Australian Postdoctoral Fellowship
from the Australian Research Council to the first author, and a Discovery Outstanding
Researcher Award from the Australian Research Council and a Wolfson Research Merit Award
from the Royal Society to the second author. The lab web address is http://www.cogsciwa.com.
Contact:
Ullrich Ecker, School of Psychology, University of Western Australia (M304), Crawley 6009,
Australia; Phone: +618 6488 3257; Fax: +618 6488 1006; E-mail: [email protected]
MISLEADING HEADLINES
2
Abstract
Information presented in news articles can be misleading without being blatantly false.
Experiment 1 examined the effects of misleading headlines that emphasize secondary content
rather than the article’s primary gist. We investigated how headlines affect readers’ processing of
factual news articles and opinion pieces, using both direct memory measures and more indirect
reasoning measures. Experiment 2 examined an even more subtle type of misdirection. We
presented articles featuring a facial image of one of the protagonists, and examined whether the
headline and opening paragraph of an article affected the impressions formed of that face even
when the person referred to in the headline was not the person portrayed. We demonstrate that
misleading headlines affect readers’ memory, their inferential reasoning and behavioral
intentions, as well as the impressions people form of faces. On a theoretical level, we argue that
these effects arise not only because headlines constrain further information processing, biasing
readers towards a specific interpretation, but also because readers struggle to update their
memory in order to correct initial misconceptions. Practical implications for news consumers and
media literacy are discussed.
Keywords: News communication; Reading comprehension; Misinformation; Memory updating;
Inferential reasoning; Facial impressions; Media literacy
MISLEADING HEADLINES
3
The Effects of Subtle Misinformation in News Headlines
Information that is initially accepted as valid but is later found to be incorrect can have a
persistent influence on people’s memory and reasoning. As recently reviewed by Lewandowsky,
Ecker, Seifert, Schwarz, and Cook (2012), corrections of invalid information are often relatively
ineffective even when people demonstrably remember the correction. As a result, people
continue to rely on retracted misinformation in their inferential reasoning. For example, when the
cause of a (fictional) fire is first presented as negligence but this information is then retracted,
people continue to make inferences in line with the retracted cause (e.g., that an insurance claim
might be rejected because of negligence). This continued reliance on misinformation occurs
despite unimpaired recall of event details and of the retraction itself (Ecker, Lewandowsky,
Swire, & D. Chang, 2011a; Johnson & Seifert, 1994). These effects have also been reported in
real-world contexts, when misconceptions regarding news events are corrected (Lewandowsky,
Stritzke, Oberauer, & Morales, 2005; Nyhan & Reifler, 2010; also see Nyhan, Reifler, Richey, &
Freed, 2014).
The dual-process theory of misinformation effects (Ecker, Lewandowsky, & Tang,
2010a; Wilson & Brekke, 1994) assumes that if misinformation represented in memory is
automatically activated in response to cues, it requires strategic processing—such as the explicit
recollection of the retraction—to avoid reliance on misinformation. Misinformation effects
therefore arise when misinformation is automatically retrieved and strategic monitoring fails
(Ecker et al., 2010a; for a review, see Lewandowsky et al., 2012).
One of the few factors known to reduce continued reliance on misinformation is the
provision of a plausible alternative account. For example, in the case of the hypothetical fire, if
the negligence retraction is accompanied by a statement that arson materials were found,
references to retracted misinformation (i.e., to negligence) are typically reduced (cf. Ecker,
MISLEADING HEADLINES
4
Lewandowsky, & Apai, 2011b; Ecker et al., 2010a; Johnson & Seifert, 1994; Seifert, 2002; also
see Kendeou, Smith, & O’Brien, 2013). Compared to a simple retraction, the additional
provision of an alternative allows the reader to meaningfully revise and correct the invalid
account—or in terms of the dual-process account, it permits recruitment of a strategically-driven
memory updating process to avoid erroneous reliance on outdated misinformation during later
reasoning.
Although providing an alternative account usually reduces misinformation effects, it does
not reliably eliminate them (cf. Ecker et al., 2011b, 2010a; Johnson & Seifert, 1994; Seifert,
2002). Updating of information in memory is error-prone and often results in incomplete
updates, meaning that corrected information can linger and continue to affect memory and
reasoning (Bjork & Bjork, 1996; Ecker, Lewandowsky, Oberauer, & Chee, 2010b; Oberauer &
Vockenberg, 2009).
An Overview of the Present Study
To date, laboratory research on misinformation has looked mainly at the effects of
misleading information that eventually turns out to be unequivocally false and that is therefore
explicitly exposed and corrected (e.g., the correction in the fire example contained an explicit
retraction of the initial account, followed by the provision of the alternative). Real-world
misinformation can likewise be clearly false, and examples include false criminal accusations
(Clow & Leach, in press; also see Ecker, Lewandowsky, Fenton, & Martin, 2014a) and the nonexistent link between childhood vaccines and autism (Ratzan, 2010).
However, misinformation in the real world is often more subtly misleading, with
misdirection resulting from information that is technically true but misleading in substance. To
illustrate: Accurate numbers or trends can be communicated in a way that makes them appear to
have more (or less) practical significance than they actually do; for example, a small or
MISLEADING HEADLINES
5
practically insignificant aspect of data can be used to argue for the existence (or non-existence)
of a larger, significant issue or trend, such as claims of global cooling by climate-change
contrarians that use only a few recent cherry-picked data points (cf. Lewandowsky, 2011).
People can be misled even when nothing inaccurate is said at all, with misdirection resulting
solely from hints, innuendo, framing, word choices, or the relative placement and ordering of
various pieces of information (for a recent news-related example, see Otieno, Spada, & Renkl,
2013). For example, a person might be associated with a certain deed or role by mere
contiguity—the image of a witness might be presented in the media in a fashion that leads
readers to believe the person is a suspect (Wemple, 2013), or a statistical graph might be coupled
with an unrelated argument in a way that fools readers into assuming that the graph supports the
argument (e.g., a statistic relating to general kidney disease in a publication on GM food safety;
Arizona Center for Advanced Medicine, 2012). In these more subtle cases of misdirection,
corrections are often incidental or implicit rather than explicit (e.g., the image or statistics graph
might have an accurate caption, but the incongruence with the accompanying information is not
explicitly addressed).
The present paper deals with these more subtle types of misinformation in two
structurally-similar experiments (see Figure 1). A first experiment examined the effects of
misleading headlines on readers’ processing of news articles under realistic conditions. We did
not use blatantly and unequivocally false headlines that misrepresent the article in all its aspects,
as such headlines should be quite uncommon and are more akin to the types of misinformation
investigated in previous research. Instead, we used headlines that were more subtly misleading
by placing emphasis on a minor aspect rather than the dominant point of the accompanying
article. By misrepresenting the gist of the article, the headline may lead to an inaccurate initial
understanding of the topic. Reading the remainder of the article can potentially rectify this by
MISLEADING HEADLINES
6
providing a more complete and accurate representation, and the dominant point of the article can
thus be regarded as alternative information that should offset the effects of the initial
misdirection. In other words, the article can serve as a correction of the misinformation contained
in a headline to the extent that people update their initial, inaccurate understanding. A second
experiment looked at the subtle misconception that can result from an incongruent pairing of a
headline with a person’s image in an article. We used article headlines that did or did not refer to
the person in the photo, assuming that people would spontaneously associate the two. The initial
impression of the person in the photo should thus be biased by the headline (e.g., whether the
headline refers to a crime victim or perpetrator), which would be inaccurate in case of a
mismatch. Again, reading the article (e.g., comparing the names and roles of persons in the
article to the photo caption) could potentially rectify this by providing a more complete and
accurate representation.
Thus, the rationale behind both experiments was that, in principle, reading an entire
article provides sufficient corrective information to amend the misinformation conveyed by a
headline that is incongruent with either the article itself or an image presented in the article.
However, given the pervasive persistence of misinformation found in previous research, our
general hypothesis was that this corrective updating should be incomplete. In other words, we
expected a measurable impact of misleading headlines. In the following section, which leads up
to Experiment 1, we specifically focus on the role of headlines and their congruence with the
accompanying article. We return to the congruence of headlines and photos in the transition to
Experiment 2.
The Role of Headlines
Headlines play a substantial role in news communication. A headline serves to
summarize the main idea of an article, it permits consumers to scan a large number of news items
MISLEADING HEADLINES
7
to get an abbreviated news update or to choose which articles to read, and it serves to grab
attention and maximize interest (Dor, 2003; Geer & Kahn, 1993; Ifantidou, 2009; van Dijk,
1988). It has been argued that many readers spend more time scanning headlines than reading
articles, as this strategy will maximize their informational gain relative to invested cognitive
effort (Dor, 2003). Headlines also permit a quick assessment of whether to expect a factual news
article or an opinion piece (Graney, 1990). They provide context, facilitating comprehension and
constraining interpretation of content based on activation of relevant background knowledge
(Bransford & Johnson, 1972; Dor, 2003; Geer & Kahn, 1993; Ifantidou, 2009; Krug, George,
Hannon, & Glover, 1989; Miller, Cohen, & Wingfield, 2006; Otero & Kintsch, 1992; Wiley &
Rayner, 2000).
A headline will therefore not only be the first (and sometimes only) element of an article
that is encoded, but it will also have an effect on what information a reader focuses on—or
ignores—when reading an article (McCrudden & Schraw, 2007; Otero & Kintsch, 1992). To the
extent that a headline activates background knowledge, it will determine how information
contained in an article is integrated with pre-existing knowledge, thereby affecting what
information a reader retains (cf. Krug et al., 1989; Rawson & Kintsch, 2002; Surber &
Schroeder, 2007; but see León, 1997). More specifically, encoding information in a particular
mindset, or with a particular schema in mind, will facilitate later retrieval of information that is
congruent with that mindset (Bransford & Johnson, 1972; Dooling & Lachman, 1971). To
achieve maximum effect, headlines are often more negative in tone than the article (ACIJ, 2011),
exaggerate article gist (Andrew, 2007), or over-emphasize conflict (Althaus, Edy, Entman, &
Phalen, 1996). For example, a study on headlines during the 2004 Canadian elections found that
across various outlets, headlines of both news and opinion pieces were less neutral (i.e., more
positive or negative, and thus less objective) than the accompanying articles (Andrew, 2007).
MISLEADING HEADLINES
8
Gricean maxims of communication imply that readers expect a headline to be relevant
and accurate in relation to the article. In most cases, this is the case; however, headlines can also
misrepresent the gist of an article and can therefore be misleading (Australian Centre for
Independent Journalism [ACIJ], 2011; Althaus, Edy, & Phalen, 2001). For example, the online
version of the UK’s Express recently featured the headline “Air pollution now leading cause of
lung cancer” (Rawle, 2013) above an article that, while stating that air pollution was “a leading
environmental cause of cancer deaths” [emphasis added], also cited a cancer researcher that
“…although air pollution increases the risk of developing lung cancer by a small amount, other
things have a much bigger effect on our risk, particularly smoking.” If a headline mismatches the
article’s content, it may affect readers’ comprehension of and memory for the article, both
because reading the headline will constrain subsequent processing of the article, and because the
headline itself provides an inaccurate summary of the article, which may be difficult to correct.
A handful of studies have investigated the role of the congruence between a news
headline and the associated article, but the results have been inconclusive. It is therefore still
unclear what the cognitive consequences of misleading headlines are. Tannenbaum (1953) paired
news articles with various headlines, each emphasizing a different aspect of the articles. He
found that headlines can influence interpretation of the articles, for example by shifting
participants’ responses regarding the guilt of a defendant in a murder trial. He also found,
however, that this effect did not occur consistently across scenarios, and that it was present
mainly after superficial reading of an article. Geer and Kahn (1993) paired an article on a
political candidate’s education policy with either a neutral, a positively slanted or negatively
slanted headline. The authors used open-ended questions to assess people’s support for the
policy, and an analog scale to measure support for the politician. They found that a negative
headline led to more negative assessment of the policy but not the political candidate, relative to
MISLEADING HEADLINES
9
a neutral or positive headline (which did not differ). Pfau (1995) presented participants with a
New York Times article concerning a riot that led to a number of police officers being injured.
The riot was described in the headline as either a “black riot” (as in the original article), or a
“union riot.” Participants rated the violence of the incident and the behavior of the involved
parties on five-point scales, and were also asked to write a pretend “letter to the editor”
concerning the article. Pfau reported that the riot was perceived as more violent when the
headline referred to a “black riot” rather than a “union riot.” Likewise, with the “black” label
participants’ letters expressed more fear-related thoughts, referred more often to the alleged
brutality of the incident, and perceived the article as more strongly biased (against the
protesters).
Other research, however, failed to find misleading-headline effects. Leventhal and Gray
(1991) presented crime articles together with headlines that were either neutral or biased in favor
of the accused or the victim. Participants were asked to recall various details regarding the
protagonists, and were also asked to assess who was responsible for the crime and how severely
they should be punished. Leventhal and Gray reported that insinuating headlines affected neither
memory for article content nor assessment of the crimes, arguing that whatever effect headlines
might have on the reader’s initial understanding will be offset and corrected by further reading.
Condit et al. (2001) used an article describing research into diabetes, emphasizing the multicausality of the disease and hence making a case against genetic determinism. The headline was
either congruent or incongruent (i.e., it conveyed a sense of low or high genetic determinism).
The authors reported that reading the article shifted participants’ beliefs in the direction
suggested by the article, but that this effect was independent of the congruence of the headline.
Follow-up interviews suggested that many participants were aware of and dissatisfied with the
misleading headline in the incongruent condition, but that they were able to discount it.
MISLEADING HEADLINES
10
The Obviousness of Misdirection
Condit et al.’s (2001) study provides a possible clue to why misleading-headline effects
are not found consistently, and will provide a starting point to derive predictions for Experiment
1. News articles with misleading headlines typically involve only incidental or implicit
corrections, leaving it up to the reader to notice the conflict between the headline and the article.
To the extent that such a conflict is not detected, there will be little incentive to engage in
strategic updating processes. In other words, one could assume that strategically applied
corrective processing during either the reading of an article or its retrieval would require that a
person notices the need for updating (cf. Figure 1; also Ecker et al., 2010a; Guzzetti, Snyder,
Glass, & Gamas, 1993; Hinze, Slaten, Horton, Jenkins, & Rapp, in press; Marsh & Fazio, 2006;
Rapp, Hinze, Kohlhepp, & Ryskin, 2014; Rapp & Kendeou, 2007, 2009; van den Broek &
Kendeou, 2008; Wilson & Brekke, 1994). In terms of the dual-process model, it follows that if
no conflict between headline and content is perceived, processing of the article will be largely
automatic and the misleading headline can ‘freely’ unfold its effect during both comprehension
and retrieval; in this case, the impact of a misleading headline should be measurable. Conversely,
if a conflict is perceived, there is at least an opportunity for the impact of the misleading headline
to be attenuated, although it remains to be seen how well people can update their memories.
We expected more obviously misleading headlines to lead to a higher rate of conflict
detection, which in turn may trigger the deployment of more strategic processing resources
towards corrective gist-memory updating—while potentially drawing resources away from the
encoding of other details (cf. Magliano, Trabasso, & Graesser, 1999; Marsh & Fazio, 2006). This
implies that a more obviously misleading headline may not affect inferential reasoning but may
affect memory for the article. By contrast, we expected less obviously misleading headlines to
have greater impact on reader’s comprehension and reasoning because they are less likely to
MISLEADING HEADLINES
11
trigger strategic corrective updating; we did not, however, expect an effect on memory for the
article because the headline should be less likely to lead to the perception of conflict and the
associated diversion of processing resources.
To manipulate the obviousness of the misdirection, we used two different types of article:
factual articles and opinion pieces. Factual articles discussed a quantitative variable (e.g., crime
rate) that exhibited an insignificant short-term trend (e.g., an increase from the previous year of
0.2%) embedded in a significant and opposing long-term trend (e.g., a continuous decline of 10%
during the last decade), and the headline focused either on the insignificant or significant trend;
thus, in the former case, the headline quite clearly contradicted the main fact discussed in the
article. Opinion pieces contrasted a layperson’s opinion with the opinion of an expert and the
headline highlighted either the layperson’s opinion or the arguably more trustworthy expert
opinion. The misdirection by the headline was less obvious than in the factual condition because
it required an assessment and weighing up of perceived expertise. We therefore assumed that the
headline-article conflict would be more obvious in a factual piece than in an opinion piece.
We employed both fact recall measures and inferential reasoning measures, and made
two predictions: Misinformation effects on inferential reasoning should be larger for opinion
pieces than factual pieces, or should only be present for opinion pieces. By contrast, recall should
be impaired more by a misleading headline in case of factual article compared to opinion pieces.
Experiment 1
Experiment 1 examined whether a misleading headline would affect readers’ memory for
the article or their inferential reasoning, using factual articles and opinion pieces. Experiment 1
thus employed a 2 (article type: factual/opinion) × 2 (headline congruence:
congruent/incongruent) within-subjects design.
MISLEADING HEADLINES
12
Method
Participants. A-priori power analyses (G*Power 3; Faul, Erdfelder, Lang, & Buchner,
2007) suggested a minimum sample size of 43 participants to detect an effect of small-tomedium size f = 0.2, at α = 0.05, 1-β = 0.8, and a moderate correlation between repeated
measures of r = 0.4. We tested 51 undergraduate students from the University of Western
Australia (44 females, 7 males; mean age was M = 20.0 [SD = 7.63] years).
Stimuli. We devised two factual articles and two opinion pieces, each with less than 400
words, and accompanying congruent and incongruent headlines (see Appendix A for all
materials). The factual-article topics were concerned with natural disaster fatality rates and
burglary rates; both articles explained that there was a slight short-term increase but a substantial
long-term decrease in rates overall; a headline emphasizing the short-term increase was
considered incongruent (e.g., ‘Number of Burglaries going up’, as opposed to the congruent
headline ‘Downward Trend in Burglary Rate’). The opinion articles were about the safety of
genetically modified (GM) foods and fluoride in drinking water; both articles contrasted a
citizen’s safety concerns with statements made by scientific bodies allaying such concerns; a
headline that emphasized the layperson’s opinion was considered incongruent (e.g., ‘Fears of
Fluoride in Drinking Water’ as opposed to the congruent headline ‘Fluoride Beneficial in
Drinking Water’). Each participant read all four articles (one per condition). The order of
conditions and the assignment of articles to congruent/incongruent headline conditions were
counterbalanced across participants.
Our headlines thus used relatively common “cherry-picking,” which can be essentially
misleading without being subject to the criticism of being manifestly false. To confirm those
aspects of our manipulation, we conducted a manipulation-check study (a) to test whether people
noticed the contradiction conveyed by the misleading headlines, and (b) to ascertain that people
MISLEADING HEADLINES
13
indeed found the incongruent headlines less appropriate than the congruent ones; that is, whether
participants could identify all misleading headlines—including those that went unnoticed on
their own—when provided with both congruent and incongruent alternatives.
We presented the materials to N = 15 students from the University of Western Australia
(11 females, 4 males; mean age M = 26.6 [SD = 6.52] years; none of them participated in the
main study). As in the main study, each participant read all four articles (one per condition, with
article-headline assignment counterbalanced). After reading each article, participants rated the
appropriateness of the headline on a 5-point scale1. The results concerning the headlineappropriateness rating are presented in Figure 2. A 2 × 2 repeated-measures ANOVA yielded the
predicted interaction, F(1,14) = 5.76 ; MSE = 1.53; p = .03; µp2 = 0.29. Participants found all
headlines appropriate (scores > 3) except the incongruent headlines of the factual pieces. There
was no difference between the congruent and incongruent versions of the opinion pieces,
F(1,14) = 1.22 ; MSE = 0.99; p = .29. After participants had read and rated all articles, we represented the articles together with both congruent and incongruent headlines and asked
participants to choose the more appropriate headline. Again, the results confirmed our
expectations: Participants selected the congruent headline 80 % of the time for the opinion pieces
and 90 % of the time for the factual pieces (the probability of choosing the congruent headline
80 % of the time or more by chance was p < .001). Having established that people found our
incongruent headlines misleading (i.e., less appropriate than the congruent headlines), but only
noticed the incongruence in case of the factual articles, we continue our description of the main
experiment.
Procedure. Participants were instructed to carefully read each article as they would be
asked to answer rating questions regarding the articles. They were presented with the articles one
at a time, and rated each article on 5 scales (e.g., asking how interesting, easy to read, and
MISLEADING HEADLINES
14
informative it was; see Appendix A for details) immediately after reading it. Participants were
told that they had two minutes to read each article and respond to the ratings (informal pilot
testing had shown this was sufficient time), but this time limit was not rigorously enforced.
Specifically, participants were presented with a booklet containing all articles (article
order was counterbalanced) and rating scales. For each article, the booklet contained three pages:
(1) a cover page, featuring only a number (1-4) and the respective article headline (in 26 pt
Times New Roman font), (2) the page featuring the article together with its headline (the article
was printed in two columns, in 13 pt Times New Roman font, and the headline appeared above
the article and spanned the entire width of the page; it was printed in 38-51 pt Times New
Roman font), and (3) a page with the five rating scales (i.e., the article was out of view when
participants responded to the scales). After each article, participants worked on an unrelated task
for about five minutes.
After reading all articles, a surprise test followed, targeting all articles in the same order
they were encoded. This involved two recall questions (one asking for a summary of each article,
one asking to state the main point of the article), a multiple-choice recognition question targeting
a factual detail unrelated to the main topic of the article and its headline, and three inference
questions. The inference questions asked participants to make a judgment regarding a trend
extrapolation (e.g., how much disaster-related fatality rates would change over the next three
years), future government spending (e.g., adjustment of funds for research into the health risks of
fluoridation), and a self-related behavioral intention assessment (e.g., willingness to pay extra for
non-GM food) on various fixed visual analog scales (e.g., -25% to +25%, 0 to 30%; see
Appendix A for details).
MISLEADING HEADLINES
15
Results
Memory scores. The ‘main point’ recall question was used to check adequate encoding.
Only four participants answered one of these ‘main point’ questions incorrectly, but all four
demonstrated some retention of content in response to other questions, so all were retained for
analysis.2 A memory measure was derived from the recall summary and the recognition item:
For each article, we selected the four main ideas contained in the article, and scored each as
recalled (1) or not recalled (0). Following precedent (e.g., Ecker et al., 2011a, 2011b, 2014a),
scoring was done by a trained scorer unfamiliar with the hypotheses; a random subset of 20
scoring sheets was additionally scored by a second scorer; inter-rater reliability was r = .91. We
then added the score on the recognition question (0/1) and divided the result by 5, yielding a
global memory score varying from 0 to 1.3
Mean memory scores across conditions are shown in Figure 3. We conducted a 2 (article
type: factual/opinion) × 2 (headline congruence: congruent/incongruent) repeated measures
ANOVA, which yielded a main effect of headline congruence, F(1,50) = 6.96; MSE = 0.021;
p = .01; µp2 = 0.12, qualified by a significant interaction, F(1,50) = 9.22; MSE = 0.020; p < .01;
µp2 = 0.16. The pattern was clear-cut: A misleading headline impaired memory for the article,
but only in case of a factual article (M = .46 [95 % CI = .41 – .52] in the incongruent and M = .58
[CI = .52 – .64] in the congruent condition; opinion article: M = .56 [CI = .51 – .61] and M = .55
[CI = .50 – .60], respectively).
Inference scores. Responses to the inference questions were standardized and averaged
to yield an inference score ranging from 0 to 1, where higher values indicate more reliance on the
article’s gist (i.e., the expert opinion). The reason for combining the heterogeneous item scores
into a global inference score was to reduce the impact of item-specific variance. This decision
was supported by the fact that for each article, the three inference questions were all positively
MISLEADING HEADLINES
16
but moderately correlated (with .04 < r < .49, and a mean correlation across articles of r = .26 for
the fact articles and r = .30 for the opinion articles; see Table 1 for details; we note that the
heterogeneity does not allow strong specific conclusions regarding, for example, behavioral
intentions; rather, the inference questions served to provide a more general and robust
assessment of topic-related inferential reasoning).
Mean inference scores across conditions are presented in Figure 4. A 2 (article type:
factual/opinion) × 2 (headline congruence: congruent/incongruent) repeated measures ANOVA
yielded a main effect of article type, F(1,50) = 31.23; MSE = 0.031; p < .001; µp2 = 0.38,
qualified by a significant interaction, F(1,50) = 4.25; MSE = 0.020; p = .04; µp2 = 0.08. A
contrast analysis confirmed that headline congruence had an effect for opinion articles (M = .54
[CI = .48 – .59] in the incongruent and M =.60 [CI = .55 – .64] in the congruent condition,
respectively), F(1,50) = 5.87; MSE = 0.016; p = .02; µp2 = 0.11, but not for factual articles, F < 1
(M =.44 [CI = .40 – .48] and M = .42 [CI = .38 – .46] in incongruent and congruent conditions,
respectively). Again, this is a clear-cut result: A misleading headline affected inferential
reasoning (e.g., behavioral intentions consistent with the gist of the article), but only in case of an
opinion article. We caution against the over-interpretation of the article type main effect. It is
known from previous research that the overall level of inferences depends strongly on the
specific questions that are asked and differs across scenarios (cf. in particular Ecker et al.,
2014a).
Discussion
Experiment 1 was designed to test how misleading headlines (a) impact on inferential
reasoning, and (b) how they affect memory for news articles. We take up both issues in turn.
Concerning (a), misleading headlines affected people’s inferential reasoning in case of
opinion but not factual articles. Factual articles were designed such that the incongruent headline
MISLEADING HEADLINES
17
would convey a relatively obvious mismatch with the gist of the article. By contrast, opinion
articles more explicitly presented two ‘sides of a story’—including the one featured in the
headline—and were thus assumed to produce less perceived inconsistency. A manipulationcheck confirmed that only the incongruent headlines of the factual articles were spontaneously
perceived as inappropriate, whereas for the opinion articles the misleading nature of a headline
became apparent only in comparison to an alternative.
The fact that misleading headlines for the opinion articles only affected inference scores,
but not memory, is thus consistent with a prediction derived from the dual-process model of the
continued influence of misinformation. The model argues that it requires strategic control
processes to avoid an impact of automatically activated misinformation: To the extent that
people become aware of an inconsistency—in this case, between the headline and the gist of the
article—the more they will strategically devote cognitive resources to the resolution of the
resulting conflict via corrective updating processes (cf. Ecker et al., 2010a; Marsh & Fazio,
2006; Rapp & Kendeou, 2007; van den Broek & Kendeou, 2008; Wilson & Brekke, 1994). This
explains why a misleading headline affected inference scores with opinion pieces but not factual
articles—only in the latter case was there a perceived inconsistency that triggered corrective
updating, which in turn reduced any potential impact of the misleading headline on subsequent
reasoning. In other words, the misleading opinion-article headlines probably unfolded their
effects without readers noticing their slant.
Our opinion pieces dealt with technologies that were described as safe by authoritative
bodies but evoked safety concerns from non-expert citizens. The impact of our manipulation was
reflected in the fact that people were more concerned with potential health risks of the
technologies if the headline expressed concern. The presentation of two contradictory views on a
topic is common media practice and reflects an emphasis on ‘balanced’ coverage of topics even
MISLEADING HEADLINES
18
when there is no underlying balance in empirical evidence (Boykoff & Boykoff, 2004; Clarke,
2008). It is left to the reader to gauge whether one of the views expressed should carry more
weight than the other, for example based on an appraisal of argument strength and perceived
source credibility (Chaiken & Maheswaran, 1994; Guillory & Geraci, 2013; Petty & Cacioppo,
1986). However, readers are likely to treat the newspaper itself as a credible source and put trust
in an editor’s decision to emphasize a particular viewpoint in a headline. Thus, readers might use
the headline’s slant as a heuristic for their own weighting of arguments, which in turn will affect
their interpretation of the article and their subsequent risk assessment. Our data showed this to be
the case.
Concerning (b)—the effect of a misleading headline on memory for the article—we found
an effect only for factual articles. This pattern is consistent with the idea outlined in the
introduction that a headline serves to guide the reader towards encoding of headline-congruent
information at the expense of headline-incongruent material (cf. Otero & Kintsch, 1992; Surber
& Schroeder, 2007). We additionally hypothesized that memorial effects would only be observed
when a detected inconsistency focuses cognitive resources on corrective updating thereby
drawing cognitive resources away from the encoding of an article (cf. Magliano et al., 1999;
Marsh & Fazio, 2006). A misleading headline thus had a direct impact on people’s memory only
when it was perceived as inappropriate, whereas no effects on memory were observed when the
headline was merely slanting interpretation of an article without being detectably at odds with its
content.4
Experiment 2
Experiment 1 investigated the effects of headlines that were incongruent with the content
of the accompanying article. Experiment 2 investigated a different, more subtle way in which a
headline can be misleading, namely when it is incongruent with an image featured in the article.
MISLEADING HEADLINES
19
Images depicting a protagonist are frequently included in news articles. Images can have a strong
and immediate influence on readers’ spontaneous inferences (Garry, Strange, Bernstein, &
Kinzett, 2007; Henkel, 2012; Strange, Garry, Bernstein, & Lindsay, 2011). Like headlines,
images can serve to attract attention and are usually processed before the full article is read,
likely in close temporal proximity with reading of the headline. We therefore assumed that
people would spontaneously infer that a photo of a person would be that of a person referred to
in the headline and the initial section of the article. Experiment 2 was designed to test people’s
updating of such spontaneous person-to-role associations in the case of incongruence; that is,
when the person in the picture later turns out not to be the person in the headline.
The conceptual and methodological structure of Experiment 2 was thus identical to that
of the first study, as shown in Figure 1, although this experiment used a very different set of
indirect measures. Participants read fictional newspaper articles about a criminal incident, which
described both a ‘good’ person (typically the victim) and a ‘bad’ person (the culprit). Each article
featured a person’s photograph; that person either corresponded to the person described in the
headline and the initial paragraph (headline congruent with the person in the photo), or it
corresponded to a person described in the latter part of the article (headline incongruent with the
person in the photo). The experiment thus used a 2 (person-in-photo: bad/good) × 2 (headline
congruence: congruent/incongruent) within-subjects design. As illustrated in Figure 1, the
pairing of headline and photo either led to a valid or biased initial representation of the personto-role association and an accurate or inaccurate initial impression of the person in the photo. In
case of a congruent headline there was no need for a corrective memory update, whereas the
incongruent headline condition required a corrective update. We note that in contrast to
Experiment 1, articles differed not only regarding the headline but also the opening paragraph;
MISLEADING HEADLINES
20
this was done to create a more natural, realistic reading experience, and to increase the
probability that the face would in fact be initially associated with the first described protagonist.
To assess whether or not people updated their memories, we used an indirect measure
involving face ratings. Previous research has established that face ratings are sensitive to prior
information about personality such as kindness and honesty (Gross & Crofton, 1977; Hassin &
Trope, 2000; Paunonen, 2006). For example, Paunonen (2006) reported that the face of a person
described as honest was subsequently judged as more kind and attractive. It follows that the same
face may be rated differently in response to our experimental manipulation. For example, if a
face is presented as belonging to a “bad” person (e.g., the headline implies that the face is of a
perpetrator), then it should be rated lower on positive attributes. Conversely, a face of a
presumed victim should be given more positive ratings.
To the extent that memory updating is imperfect, a face initially thought to belong to a
bad person should receive a relatively negative rating even when the face turns out to belong to a
good person and vice versa. Based on the previous relevant literature, we selected four traits that
have been described as central to the process of impression-formation: attractiveness,
trustworthiness, dominance, and aggressiveness. Attractiveness and trustworthiness typically
load high on a “valence” factor, while dominance and aggressiveness load highly on a
“dominance” factor (cf. Oosterhof & Todorov, 2008; Todorov, Said, & Verosky, 2011). We
expected an effect to most likely emerge on the valence dimension, both because the
trustworthiness trait seems most central in the present distinction between criminals and noncriminals, and because previous research reported effects on valence ratings (Gross & Crofton,
1977; Paunonen, 2006).
MISLEADING HEADLINES
21
Method
Participants. Following a-priori power analysis (see Experiment 1), we tested 47
undergraduates from the University of Western Australia (39 females, 8 males, mean age
M = 18.6 [SD = 2.9] years) via an online survey.
Stimuli. We used naturalistic face images typical of those found in newspaper articles.
Face stimuli were taken from the freely available database “Labeled Faces in the Wild” (Huang,
Ramesh, Berg, & Learned-Miller, 2007). This database features naturalistic face images
collected from the internet. A pool of 45 face images of not immediately recognizable persons
were pre-selected; selected faces had a neutral or moderately happy expression as typically seen
in newspaper-article photos, and images contained minimal background information. Each face
image was converted to a monochrome version, and cropped and re-sized (205 × 210 pixels; size
on screen approx. 6 × 6 cm) to achieve consistency across images.
The initial pool of 45 face stimuli was used in a pilot study to select the stimulus set for
the main study—16 pairs of similar, age- and gender-matched faces.. Twelve students from the
University of Western Australia, none of whom participated in the main experiment (3 males and
9 females; mean age 24.7 years), rated all faces via an online survey. Each face was rated
individually on the four traits attractiveness, dominance, aggressiveness, and trustworthiness, as
well as masculinity/femininity5 and presumed socio-economic status on a scale from 0 (low) to
10 (high). Participants also rated the expressed emotion on a scale from 0 (negative—sad, angry,
or upset) to 10 (positive—happy, loving, or joyful). Finally, they were asked to estimate the
person’s age, and whether they recognised the person (if they thought they recognized the
person, they were asked to provide a name or any person-related information that came to mind).
No participant recognized any of the depicted persons.
MISLEADING HEADLINES
22
Using the pool of 45 faces, we created 30 candidate pairs of similar faces, based on
gender, apparent age, and matching features (some faces were part of multiple pairs). We then
selected the 16 pairs with the highest similarity as determined by the pilot-study ratings (with the
constraint that each face was used in only one pair). A total of 30 × 8 t-tests on the face ratings
were carried out, with an alpha level set at α = 0.001. There were no significant differences
between the two faces of each selected pair except five pairs that differed in estimated age (by an
average margin of 14.8 years), one pair that differed in estimated socio-economic status
(difference of 1.42 on ten-point scale), and one pair that differed in rated emotionality (difference
2.33 on ten-point scale).
We then devised 16 scenarios, each involving a crime or accident. Each scenario was
associated with a specific pair of faces. For each scenario, we generated 2 news articles that were
substantially identical but manipulated the congruence of the picture. In both articles, there was
a ‘good’ person (a victim or police officer or prosecutor) and a ‘bad’ person (the culprit). The
two persons were always described as people of the same gender and similar age. To manipulate
the congruence variable, the two articles (1) altered the order in which they focused on the good
and the bad person, and (2) they had different headlines such that the headline always matched
the initial paragraph.5 For example, the headline of one article was “Man charged over Thornlie
murder”6; this article began with a description of a pensioner being arrested for murdering his
neighbor. The corresponding second article had the headline “Grandfather killed in Thornlie”,
and accordingly first described the discovery of the victim’s body before reporting the arrest of
the pensioner. Each article was presented with the image of a face (i.e., one of the two faces
associated with the particular scenario), much like a real-world newspaper article would
incorporate an image. The fictitious name of one of the two protagonists (i.e., the name of the
good or bad person) was printed beneath the face image. This resulted in 8 different versions of
MISLEADING HEADLINES
23
each of the 16 scenarios (2 headlines/articles × 2 faces × 2 names/roles; see Appendix B for
examples). The articles, the assignment of scenarios to conditions, as well as the assignment of
specific faces to good/bad roles was counterbalanced across participants.
Procedure. Participants were presented with 16 news articles (4 per condition), one at a
time, each paired with one face. Participants were instructed to read the articles carefully as they
would be asked questions about the events described and the people involved; there were no time
limits. After reading each article, participants answered two 4-alternative multiple choice
questions, which only served to check adequate encoding (e.g., “Where was the murderer
arrested?”—at the park/workplace/police station/his property). After completing all 16 news
articles, all presented faces were presented again, one-by-one and in random order, and
participants rated each face on the two valence and two dominance traits (on 0-10 scales). Lastly,
all faces were presented a third time in random order, and participants were asked to classify
each face as belonging to a ‘good’ (innocent victim, police officer, prosecutor) or ‘bad’ (culprit,
criminal) person, to test participants’ memory for the face-role associations.
Results
Comprehension scores. All participants demonstrated above-chance comprehension on
the 32 multiple-choice comprehension questions (M = 27.13; SD = 3.47; range = 18-32; the
probability of correctly guessing  18 responses is p < .0002).
Face ratings. For each participant and condition, we averaged the two valence and the
two dominance trait ratings separately and mapped them onto a 0-1 scale. Two 2 (person-inphoto: bad/good) × 2 (headline congruence: congruent/incongruent) repeated measures
ANOVAs were conducted, with significance level set at α = 0.025 to correct for multiple tests. In
these analyses, perfect memory updating during encoding would imply a main effect of personin-photo but no main effect of headline congruence and no interaction—face ratings should be
MISLEADING HEADLINES
24
based purely on whether the face depicts a good or bad person, irrespective of who people
initially thought the person in the picture might be. In contrast, the failure of updating would
imply an interaction but no main effects—face ratings should be determined entirely by the
headline’s valence (e.g., positive face ratings in case the headline focus was on the ‘good’
person, and negative ratings when the headline focused on the ‘bad’ person). Because both
headlines could be either congruently paired with a good (bad) person’s photo, or incongruently
paired with a culprit’s (good person’s) photo, reliance on the headline alone would result in an
interaction.
The analysis on dominance ratings yielded no significant effects, all F(1,46) < 3.43, all
p > .07. The analysis on valence ratings yielded a main effect of person-in-photo,
F(1,46) = 11.31; MSE = 0.006; p = .002; η2p = .20, which was qualified by an interaction,
F(1,46) = 12.25; MSE = 0.017; p = .001; η2p = .21 (see Figure 5). When the headline was
congruent with the image (i.e., no corrective updating required), a bad person’s image was rated
more negatively than a good person’s image (M = .32 [CI = .27 – .36] for the bad person and
M = .42 [CI = .39 – .46] for the good person; compare the two dark bars in Figure 5),
F(1,46) = 25.83; MSE = 0.010; p < .001; η2p = .36. This effect was eliminated (and even reversed
numerically) if headline and photo mismatched, and the face ratings shifted in the direction
suggested by the headline: the good person’s image was rated more negatively with an
incongruent headline that focused on the culprit (M = .35 [CI = .30 – .40]), compared to the
congruent case (compare the two rightmost bars in Figure 5), F(1,46) = 10.03; MSE = 0.012;
p = .003; η2p = .18, and the bad person’s image was rated more positively with an incongruent
headline that focused on the good person (M = .38 [CI = .34 – .43]), compared to the congruent
case (compare the two leftmost bars), F(1,46) = 6.92; MSE = 0.014; p = .01; η2p = .13. In other
words, the misleading headline pulled people’s face perceptions away from the true role of the
MISLEADING HEADLINES
25
person pictured; when headline and identity matched (dark bars), ratings were extreme, when
there was incongruence (light bars), ratings were intermediate, thus reflecting partial updating.7
Memory for face-role associations. Memory for face-role associations was relatively
poor overall (M = .60; CI = .56 – .65). Only a subset of 17 participants demonstrated abovechance memory for the correct (updated) face-role associations, using a lenient criterion (i.e.,
role-memory > .69; which maps into a probability of guessing ≥ 11 out of 16 correctly of
p = .11). It might be expected that the face-rating interaction of person-in-photo and headline
congruence might become less prominent, and the main effect of person-in-photo more
prominent, in this subgroup of participants because if memory updating was successful the
interaction with a main effect of person-in-photo should disappear. However, the interaction was
still significant in this subgroup, F(1,16) = 20.83, MSE = 0.021; p < .001; η2p = .57, but not the
main effects (Fs < 1).
Discussion
Experiment 2 showed that valence ratings of faces were affected not only by the actual
role of the pictured person, but also by the congruence of the headline plus opening paragraph
and the photo (in the remainder of this discussion section, the term ‘headline’ will refer to the
larger frame conveyed by the headline and the opening paragraph together). Before discussing
this main finding, we acknowledge that the effect was limited to the valence dimension;
regarding the dominance dimension, we speculate that some of the ‘good’ people in our stories
were police officers and prosecutors, and might have been perceived as rather dominant,
compared to a victim or even some of the ‘bad’ people in our scenarios (this notion is
corroborated by the fact that the dominance trait was the only trait out of the four we used [i.e.,
attractiveness, trustworthiness, dominance, aggressiveness] not to exhibit an interaction between
MISLEADING HEADLINES
26
person-in-photo and headline congruence). In the context of crime scenarios, at least, the valence
dimension may thus be the primary dimension of classification for news readers.
Results from the valence ratings demonstrated that readers spontaneously integrated
information from a headline with a photo in the associated article. The congruence of headline
and photo therefore affected subsequent face valence ratings: Headlines that were incongruent
with the photo led people to make face judgments that were more in line with the headline, even
though the article later on clarified who the person in the picture was. The presence of the
significant interaction demonstrated that there was some updating of the initial face-role
assumptions, but that it was only partially effective. In other words, if a reader initially assumed
the face belonged to the culprit (victim), they then continued to rate the face negatively
(positively) even when the face actually belonged to an innocent victim (culprit).
This pattern of results is in line with research on the continued influence of
misinformation, where more direct and explicit corrections are likewise often found to reduce but
not eliminate the influence of misinformation on memory and inferential reasoning (Ecker et al.,
2010a, 2011a, 2011b, 2014a; Johnson & Seifert, 1994; Lewandowsky et al., 2012; Seifert, 2002).
The results also support the notion that misinformation effects can arise from partial memory
updating, where attempted correction of irrelevant information, or its removal from memory, is
often incomplete, leading to lingering representations of outdated information (e.g., Oberauer &
Vockenberg, 2009).
The notion of partial updating is further supported by our finding that the subgroup of
participants who demonstrated superior updating in their explicit memory for face-role
associations did not demonstrate superior updating in their face ratings, compared to the overall
group. In other words, people with relatively good memory for the actual role of the person
pictured failed to demonstrate corrective updating in their face ratings: people rated a face they
MISLEADING HEADLINES
27
initially associated with a negative (positive) role more negatively (positively), even when they
subsequently updated the role-association in memory.
In sum, the results of Experiment 2 mesh well with those of Experiment 1 (in particular
those of the opinion article conditions): Reading a news article that disconfirms an initial,
incorrect conception is not sufficient to remediate the impact of a misleading headline (cf.
Figure 1). Both experiments thus provided evidence for a continued influence effect of subtle
misinformation conveyed by misleading news headlines.
General Discussion
There can be little doubt that misleading headlines result in misconceptions in readers
who do not read beyond the headlines (e.g., Wegner, Wenzlaff, Kerker, & Beattie, 1981).
However, it was hitherto unclear what the effects of misleading headlines are if the entire article
is read, thus providing an opportunity for any initial misconceptions to be corrected (cf. Condit et
al., 2001; Leventhal & Gray, 1991; Tannenbaum, 1953). The present research suggests that
misleading headlines affect readers’ memory for news articles or their inferential reasoning, and
even their impressions of faces featured in the articles.
The effects of misleading headlines are likely to arise concurrently from multiple
cognitive mechanisms. First, readers use available information to constrain further information
processing. This means that any incoming evidence will always be weighted and interpreted in
light of information already received, and a headline can thus serve to bias processing towards or
away from a specific interpretation (Dor, 2003; Ifantidou, 2009; Krug et al., 1989; McCrudden &
Shaw, 2007; Miller et al., 2006; Otero & Kintsch, 1992; Surber & Schroeder, 2007).
Second, correcting the misinformation conveyed by a misleading headline is a difficult
task. Particularly in cases of non-obvious misdirection, readers may not be aware of an
inconsistency, and may thus not initiate any corrective updating. By contrast, if a headline is
MISLEADING HEADLINES
28
perceived as inappropriate, people may be able to correct its influence on their understanding of
the article, although this correctional effort itself may withdraw resources from mnemonic
processing of the article and may thus impair memory. In sum, these effects further corroborate
the notion that misinformation tends to influence people’s memory and reasoning continuously
despite corrections (cf. Lewandowsky et al., 2012). Such effects are likely caused by a failure of
strategic monitoring and memory updating processes (Ecker et al., 2010a). A misleading
headline can thus do damage despite genuine attempts to accurately comprehend an article. As a
certain motivation and effort is required to engage in such strategic processing, headline effects
may be also affected by the reader’s pre-existing attitudes (cf. Ecker et al., 2014a; Nyhan &
Reifler, 2010). Future research should thus aim to investigate the interaction of pre-existing
attitudes and headline slant, using a wider variety of articles.
The present paper showed that misleading headlines can lead to misconceptions and
misinformed behavioral intentions. The practical implications of this research are clear: News
consumers must be (made) aware that editors can strategically use headlines to effectively sway
public opinion and influence individuals’ behavior (cf. ACIJ, 2011). More specifically, the
present research suggests the following: (1) It may be the more indiscernibly misleading
headlines in particular, merely highlighting one opinion over another, that have an impact on the
reader’s reasoning. This will occur even though—or as we have argued: precisely because—
readers may not spontaneously perceive the headline as misleading. Such rather indiscernibly
misleading headlines can be seen as an extension of the ‘balanced evidence’ approach so popular
in modern-day journalism (Boykoff & Boykoff, 2004; Clarke, 2008): not only do many articles
give equal space to the evidence-based opinion of an expert and the unsubstantiated opinion of a
non-expert, but the non-expert’s opinion can be further validated by also being highlighted in the
headline.
MISLEADING HEADLINES
29
(2) Using a headline to emphasize such a non-expert view can be utilized as a tool to
influence beliefs and attitudes—in the present study, to fuel safety concerns regarding
technologies believed to be safe—and reduce people’s willingness to follow expert advice (also
see Otieno et al., 2013). In our specific case, a headline amplifying the opinion of a non-expert
citizen made readers more concerned about the safety of GM food and water fluoridation, despite
receiving the identical information containing expert advice arguing for the safety of these
practices. This is a worrying implication from our research, which is particularly concerning
given the generally decreasing trust in scientific experts among segments of the public (Gauchat,
2012). Accordingly, misinformation in the media has, for example, been linked to declining
vaccination rates (Smith, Yarwood, & Salisbury, 2007; also see Health Protection Agency, 2011)
and rejection of climate-change mitigation policies (Ding, Maibach, Zhao, Roser-Renouf, &
Leiserowitz, 2011), and can potentially play a role in determining the support (or non-support)
for any matter of public interest, from the invasion of a foreign country (Lewandowsky, Stritzke,
Freund, Oberauer, & Krueger, 2013; Lewandowsky, Stritzke, Oberauer, & Morales, 2009) to the
rejection of asylum seekers (Macken-Horarik, 2003; Pedersen, Clarke, Dudgeon, & Griffiths,
2005).
(3) Finally, a headline can be used to cast someone in a dubious light even when every
word in both the headline and the accompanying article is accurate. The impression of a person
can be affected merely by placing someone’s face next to an insinuating headline and structuring
the accompanying article in a way that implicates the person in the photo was involved in
misdemeanor. To the best of our knowledge, this is the first demonstration of such an effect in
the context of news articles. It mimics a similar effect discussed in media psychology, where
advertisements placed strategically next to relevant editorial content are assumed to have
stronger impact on consumers (cf. Baerns, 2003). In this area, laws regarding the ‘principle of
MISLEADING HEADLINES
30
separation’ (i.e., clear separation of advertisements and editorial content) are in place in many
countries to protect consumers from being misled. Our research shows that this principle should
also be considered and applied to news articles and the images they contain.
As knowledge about misinformation effects can reduce their impact (Ecker et al., 2010a),
awareness of the effects of misleading headlines should thus be considered a critical aspect of
media literacy and media education. The more people are aware of the issue, the more will they
skeptically assess headlines and notice misleading ones, and the more likely will they engage in
strategic processing to reduce their impact. Moreover, the modern-day fact-checking movement,
which has allowed bloggers with little or no research budget to discuss the veracity of claims
made by public figures such as politicians or industry spokespersons (cf. Dobbs, 2012), should
not only focus on debunking misrepresentations but should also pay attention to how factually
accurate contents are framed by the editors of newspapers and news websites.
MISLEADING HEADLINES
31
References
Allport, F. H., & Lepkin, M. (1943). Building war morale with news-headlines. Public Opinion
Quarterly, 7, 211-221.
Althaus, S. L., Edy, J. A., & Phalen, P. F. (2001). Using substitutes for full-text news stories in
content analysis: which text is best? American Journal of Political Science, 45, 707-723.
Althaus, S. L., Edy, J. A., Entman, R. M., & Phalen, P. F. (1996). Revising the indexing
hypothesis: Officials, media, and the Libya crisis. Political Communication, 13, 407-421.
Andrew, B. C. (2007). Media-generated shortcuts: Do newspaper headlines present another
roadblock for low-information rationality? The Harvard International Journal of
Press/Politics, 12, 24-43.
Australian Centre for Independent Journalism (2011). A sceptical climate. Media coverage of
climate change in Australia 2011: Part 1—Climate change policy. University of
Technology, Sydney. Available from http://www.acij.uts.edu.au/pdfs/sceptical-climatepart1.pdf
Arizona Center for Advanced Medicine (2012). GMO foods—Are they safe? Look at the
evidence. Retrieved from http://arizonaadvancedmedicine.com/gmo-foods-are-they-safelook-at-the-evidence/
Baerns, B. (2003). Separating advertising from programme content: The principle and its
relevance in communications practice. Journal of Communication Management, 8, 101112.
Bjork, E.L., & Bjork, R.A. (1996). Continuing influences of to-be-forgotten information.
Consciousness and Cognition, 5, 176-196.
Boykoff, M. T., & Boykoff, J. M. (2004). Balance as bias: Global warming and the US prestige
press. Global Environmental Change, 14, 125-136.
MISLEADING HEADLINES
32
Bransford, J. D., & Johnson, M. K. (1972). Contextual prerequisites for understanding: Some
investigations of comprehension and recall. Journal of Verbal Learning and Verbal
Behavior, 11, 717-726.
Chaiken, S., & Maheswaran, D. (1994). Heuristic processing can bias systematic processing:
Effects of source credibility, argument ambiguity, and task importance on attitude
judgment. Journal of Personality and Social Psychology, 66, 460-473.
Clarke, C. (2008). A question of balance: The autism-vaccine controversy in the British and
American elite press. Science Communication, 30, 77-107.
Clow, K. A., & Leach, A.-M. (in press). After innocence: Perceptions of individuals who have
been wrongfully convicted. Legal and Criminological Psychology.
Condit, C.M., Ferguson, A., Kassel, R., Thadhani, C., Gooding, H.C., Parrott, R., … Jones, C.
(2001). An exploratory study of the impact of news headlines on genetic determinism.
Science Communication, 22, 379-395.
Ding, D., Maibach, E. W., Zhao, X., Roser-Renouf, C., & Leiserowitz, A. (2011). Support for
climate policy and societal action are linked to perceptions about scientific agreement.
Nature Climate Change, 1, 462-466.
Dobbs, M. (2012). The rise of political fact-checking—How Reagan inspired a journalistic
movement: A reporter’s eye view. New America Foundation. Available online:
http://newamerica.net/sites/newamerica.net/files/policydocs/The_Rise_of_Political_Factchecking.pdf.pdf
Dooling, D. J., & Lachman, R. (1971). Effects of comprehension on retention of prose. Journal
of Experimental Psychology, 88, 216-222.
Dor, D. (2003). On newspaper headlines as relevance optimizers. Journal of Pragmatics, 35,
695-721.
MISLEADING HEADLINES
33
Ecker, U. K. H., Lewandowsky, S., & Apai, J. (2011b). Terrorists brought down the plane! - No,
actually it was a technical fault: Processing corrections of emotive information. Quarterly
Journal of Experimental Psychology, 64, 283-310.
Ecker, U. K. H., Lewandowsky, S., Fenton, O., & Martin, K. (2014a). Do people keep believing
because they want to? Pre-existing attitudes and the continued influence of
misinformation. Memory & Cognition, 42, 292-304.
Ecker, U. K. H., Lewandowsky, S., Oberauer, K., & Chee, A. E. H. (2010b). The components of
working memory updating: An experimental decomposition and individual differences.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 36, 170-189.
Ecker, U. K. H., Lewandowsky, S., & Tang, D. T. W. (2010a). Explicit warnings reduce but do
not eliminate the continued influence of misinformation. Memory & Cognition, 38, 10871100.
Ecker, U. K. H., Lewandowsky, S., Swire, B., & Chang, D. (2011a). Correcting false information
in memory: Manipulating the strength of misinformation encoding and its retraction.
Psychonomic Bulletin & Review, 18, 570-578.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical
power analysis for the social, behavioral, and biomedical sciences. Behavior Research
Methods, 39, 175-191.
Garry, M., Strange, D., Bernstein, D. M., & Kinzett, T. (2007). Photographs can distort memory
for the news. Applied Cognitive Psychology, 21, 995-1004.
Gauchat, G. (2012). Politicization of science in the public sphere: A study of public trust in the
United States, 1974 to 2010. American Sociological Review, 77, 167-187.
Geer, J. G., & Kahn, K. F. (1993). Grabbing attention: An experimental investigation of
headlines during campaigns. Political Communication, 10, 175-91.
MISLEADING HEADLINES
34
Graney, J. M. (1990). Determination of fact and opinion: A critical reading problem. Journal of
Psycholinguistic Research, 19, 147-166.
Gross, A. E., & Crofton, C. (1977). What is good is beautiful. Sociometry, 40, 85-90.
Guillory, J. J., & Geraci, L. (2013). Correcting erroneous inferences in memory: The role of
source credibility. Journal of Applied Research in Memory and Cognition, 2, 201-209.
Guzzetti, B. J., Snyder, T. E., Glass, G. V., & Gamas, W. S. (1993). Promoting conceptual
change in science: A comparative meta-analysis of instructional interventions from
reading education and science education. Reading Research Quarterly, 28, 117–159.
Johnson, H. M., & Seifert, C. M. (1994). Sources of the continued influence effect: When
misinformation in memory affects later inferences. Journal of Experimental Psychology:
Learning Memory and Cognition, 20, 1420-1436.
Hassin, R., & Trope, Y. (2000). Facing faces: Studies on the cognitive aspects of physiognomy.
Journal of Personality and Social Psychology, 78, 837-852.
Health Protection Agency (2011, 21 October). Completed primary courses at two years of age:
England and Wales, 1966-1977, England only after 1978 onwards. Retrieved from
http://www.hpa.org.uk/web/HPAweb&HPAwebStandard/HPAweb_C/1195733819251
Henkel, L. A. (2012). Seeing photos makes us read between the lines: The influence of photos on
memory for inferences. Quarterly Journal of Experimental Psychology, 65, 773-795.
Huang, G. B., Ramesh, M., Berg, T., & Learned-Miller, E. (2007). Labeled Faces in the Wild: A
database for studying face recognition in unconstrained environments. University of
Massachusetts, Amherst, Technical Report 07-49. Available online from http://viswww.cs.umass.edu/lfw/index.html.
Ifantidou, E. (2009). Newspaper headlines and relevance: Ad hoc concepts in ad hoc contexts.
Journal of Pragmatics, 41, 699-720.
MISLEADING HEADLINES
35
Kendeou, P., Smith, E. R., & O'Brien, E. J. (2013). Updating during reading comprehension:
Why causality matters. Journal of Experimental Psychology: Learning Memory and
Cognition, 39, 854-865.
Krug, D., George, B., Hannon, S. A., & Glover, J. A. (1989). The effect of outlines and headings
on readers’ recall of text. Contemporary Educational Psychology, 14, 111-123.
León, J.A. (1997). The effects of headlines and summaries on news comprehension and recall.
Reading and Writing, 9, 85-106.
Leventhal, G., & Gray, S. J. (1991). Can innuendos in headlines affect perceptions?
Psychological Reports, 69, 801-802.
Lewandowsky, S. (2011). Popular Consensus Climate Change Is Set to Continue. Psychological
Science, 22, 460-463.
Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N., & Cook, J. (2012).
Misinformation and its correction: Continued influence and successful debiasing.
Psychological Science in the Public Interest, 13, 106-131.
Lewandowsky, S., Stritzke, W. G. K., Freund, A. M., Oberauer, K., & Krueger, J. I. (2013).
Misinformation, disinformation, and violent conflict: From Iraq and the "War on Terror"
to future threats to peace. American Psychologist, 68, 487-501.
Lewandowsky, S., Stritzke, W. G. K., Oberauer, K., & Morales, M. (2005). Memory for fact,
fiction, and misinformation: The Iraq War 2003. Psychological Science, 16, 190-195.
Lewandowsky, S., Stritzke, W. G. K., Oberauer, K., & Morales, M. (2009). Misinformation and
the "War on Terror": When memory turns fiction into fact. In W. G. K. Stritzke, S.
Lewandowsky, D. Denemark, J. Clare, & F. Morgan (Eds.), Terrorism and Torture: An
Interdisciplinary Perspective (pp. 179-203).
MISLEADING HEADLINES
36
Marsh, E. J., & Fazio, L. K. (2006). Learning errors from fiction: Difficulties in reducing
reliance on fictional stories. Memory & Cognition, 34, 1140-1149.
Macken-Horarik, M. (2003). Working the borders in racist discourse: The challenge of the
‘Children Overboard Affair’ in news media texts. Social Semiotics, 13, 283-303.
Magliano, J. P., Trabasso, T., & Graesser, A. C. (1999). Strategic Processing during
Comprehension. Journal of Educational Psychology, 91, 615-629.
McCrudden, M. T., & Schraw, G. (2007). Relevance and goal-focusing in text processing.
Educational Psychology Review, 19, 113-139.
Miller, L. M. S., Cohen, J. A., & Wingfield, A. (2006). Contextual knowledge reduces demands
on working memory during reading. Memory & Cognition, 34, 1355-1367.
Nyhan, B., & Reifler, J. (2010). When corrections fail: The persistence of political
misperceptions. Political Behavior, 32, 303-330.
Nyhan, B., Reifler, J., Richey, S., & Freed, G. L. (2014). Effective messages in vaccine
promotion: A randomized trial. Pediatrics, 133, 835-842.
Oberauer, K., & Vockenberg, K. (2009). Updating of working memory: Lingering bindings.
Quarterly Journal of Experimental Psychology, 62, 967-987.
Oosterhof, N. N., & Todorov, A. (2008). The functional basis of face evaluation. Proceedings of
the National Academy of Sciences of the United States of America, 105, 11087-11092.
Otero, J., & Kintsch, W. (1992). Failures to detect contradictions in a text: What readers believe
versus what they read. Psychological Science, 3, 229-235.
Otieno, C., Spada, H., & Renkl, A. (2013). Effects of news frames on perceived risk, emotions,
and learning. PLoS ONE, 8, e79696.
Paunonen, S. V. (2006). You are honest, therefore I like you and find you attractive. Journal of
Research in Personality, 40, 237-249.
MISLEADING HEADLINES
37
Pedersen, A., Clarke, S., Dudgeon, P., & Griffiths, B. (2005). Attitudes toward Indigenous
Australians and asylum seekers: The role of false beliefs and other social-psychological
variables. Australian Psychologist, 40, 170-178.
Perfetti, C.A., Beverly, S., Bell, L., Rodgers, K., & Faux, R. (1987). Comprehending newspaper
headlines. Journal of Memory and Language, 26, 692-713.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances
in Experimental Social Psychology, 19, 123-205.
Pfau, M. R. (1995). Covering urban unrest: The headline says it all. Journal of Urban Affairs, 17,
131-41.
Rapp, D. N., Hinze, S. R., Kohlhepp, K., & Ryskin, R. A. (2014). Reducing reliance on
inaccurate information. Memory & Cognition, 42, 11-26.
Rapp, D. N., & Kendeou, P. (2007). Revisiting what readers know: Updating text representations
during narrative comprehension. Memory & Cognition, 35, 2019-2032.
Rapp, D. N., & Kendeou, P. (2009). Noticing and revising discrepancies as texts unfold.
Discourse Processes, 46, 1-24.
Ratzan, S. C. (2010). Editorial: Setting the record straight: Vaccines, autism, and The Lancet.
Journal of Health Communication, 15, 237-239.
Rawle, T. (2013, October 17). Air pollution now leading cause of lung cancer. Express.
Retrieved from http://www.express.co.uk/news/health/437473/Air-pollution-nowleading-cause-of-lung-cancer
Rawson, K. A., & Kintsch, W. (2002). How does background information improve memory for
text content? Memory & Cognition, 30, 768-778.
Seifert, C. M. (2002). The continued influence of misinformation in memory: What makes a
correction effective? Psychology of Learning and Motivation, 41, 265-292.
MISLEADING HEADLINES
38
Smith, A., Yarwood, J., & Salisbury, D. M. (2007). Tracking mothers' attitudes to MMR
immunisation 1996-2006. Vaccine, 25, 3996-4002.
Strange, D., Garry, M., Bernstein, D. M., & Lindsay, D. S. (2011). Photographs cause false
memories for the news. Acta Psychologica, 136, 90-94.
Surber, J. R., & Schroeder, M. (2007). Effect of prior domain knowledge and headings on
processing of informative text. Contemporary Educational Psychology, 32, 485-498.
Tannenbaum, P. H. (1953). The effect of headlines on the interpretation of news stories.
Journalism Quarterly, 30, 189-97.
Todorov, A., Said, C. C., & Verosky, S. C. (2011). Personality impressions from facial
appearance. In A. J. Calder, G. Rhodes, M. H. Johnson, & J. V. Haxby (Eds.), The
Oxford Handbook of Face Perception (pp.631-651). Oxford: Oxford University Press.
van den Broek, P., & Kendeou, P. (2008). Cognitive processes in comprehension of science
texts: The role of co-activation in confronting misconceptions. Applied Cognitive
Psychology, 22, 335-351.
van Dijk, T. A. (1988). News as Discourse. Hillsdale, NJ: Lawrence Erlbaum Associates
Wegner, D. M., Wenzlaff, R., Kerker, R. M., & Beattie, A. E. (1981). Incrimination through
innuendo: Can media questions become public answers? Journal of Personality and
Social Psychology, 40, 822-832.
Wemple, E. (2013). New York Post: ‘Bag Men’ headline was an ‘attention-getter’. The
Washington Post, November 4, 2013. Retrieved from:
http://www.washingtonpost.com/blogs/erik-wemple/wp/2013/11/04/new-york-post-bagmen-headline-was-an-attention-getter/
Wiley, J., & Rayner, K. (2000). Effects of titles on the processing of text and lexically
ambiguous words: Evidence from eye movements. Memory & Cognition, 28, 1011-1021.
MISLEADING HEADLINES
Wilson, T. D., & Brekke, N. (1994). Mental contamination and mental correction: Unwanted
influences on judgments and evaluations. Psychological Bulletin, 116, 117-142.
39
MISLEADING HEADLINES
40
Footnotes
1
Participants also provided ratings on five other scales, which only served as distractors
and are thus not relevant here—see Procedure of main experiment and Appendix A for details.
2
Their exclusion did not affect the outcome.
3
Recall and recognition scores were combined to obtain a general memory index.
However, because recall targeted the entire article while the recognition question targeted an
arbitrary detail, we also analyzed only the recall summary score separately; both analyses yielded
comparable results so we report the combined analysis in the following (the recognition score by
itself showed a similar effect pattern also).
4
Inspection of responses suggests that lower memory scores were mostly due to omission
errors, although there were also commission errors (i.e., misrepresentations). Alas, the available
data do not permit a more fine-grained analysis.
5
Male faces were rated on masculinity, female faces on femininity. This scale was also
included in the main study, but it was dropped from analyses because it was unclear how to
combine ratings for male and female faces onto the same scale; that is, it was impossible to
ascertain whether or not a specific masculinity rating of a male face (e.g., x = 3) was equivalent
to the analogous femininity rating of a female face (i.e., 10 – x = 7). Moreover, although
Todorov, Said, and Verosky (2011) mentioned masculinity as related to the ‘dominance’ factor,
masculinity was not included in the original factor analyses of Oosterhof and Todorov (2008).
6
Thornlie is a Perth suburb.
7
We note that there are alternative ways to present the data in Figure 5. Specifically, the
analysis could have included a factor of “headline focus”, that is, whether the headline focused
on a good or bad person. Such an analysis would return a main effect of headline focus such that
a person ostensibly referred to as a good person in the headline is rated more positively than a
MISLEADING HEADLINES
41
person referred to as a bad person (compare the two outer bars in Figure 5 with the two middle
bars). This alternative way to look at the data does not change the general interpretation of
results but does make clear that readers initially associated the person in the photo with the role
mentioned in the headline (an association we did not measure directly).
MISLEADING HEADLINES
42
Appendix A—Materials used in Experiment 1 (to be placed in Online Supplement)
News Articles (with Incongruent / Congruent Headline)
Factual articles.
Number of Burglaries Going Up (incongruent) / Downward Trend in Burglary Rate
(congruent)
The Australian Office of Crime Statistics
trend in burglary rates has actually been
(AOCS) released their annual crime figures
downward for over 10 years, and burglaries
yesterday. The report shows that residential
have declined by 15% since 2000. So
burglary rates in Australia have gone up
inducing panic in communities is
0.2% since last year.
unnecessary.”
According to Ted Forest from WA’s
Across the states in Australia, households
Neighbourhood Watch, “most burglaries are
in the Northern Territory and Western
committed by teenagers who see an
Australia were most likely to be the victims
opportunity. We strongly urge homeowners
of a break-in (7.7% and 5.1% of households
to lock their doors and windows, install an
respectively),
whilst
households
in
alarm, and make arrangements with their
Tasmania and Victoria were the least likely
neighbours when they go away to watch
to be targeted (2.6% and 2.8% of
over their property. People should also avoid
households, respectively).
leaving expensive items in view of others,
Household
victimization
surveys,
especially during school holidays, which is a
conducted across OECD countries in 2005,
popular time with burglars.”
found that Australia had the 9th highest
Frank Whitacker from the AOCS pointed
victimization rate for burglaries (3.4%),
out that the “numerical increase in burglaries
which was higher than Japan, Italy, and
is not a significant increase by any means. If
Canada, but lower than the United Kingdom,
you look at the bigger picture, the overall
New Zealand, and the United States.
More Natural Disasters, More Deaths (incongruent) / More Natural Disasters, Fewer
Deaths (congruent)
Natural disasters are on the rise. About
Earth might seem like a more dangerous
two-thirds of the increase in the number of
place than ever, given the constant media
natural disasters is the result of rises in
reports of natural disasters. At a press
hydro-meteorological
disasters.
These
conference yesterday, the director of the
disasters include droughts, hurricanes,
Center for Research on the Epidemiology of
typhoons and floods and have been
Disasters (CRED) stated “it is logical to
increasing over the past 25 years. In 1980,
think that the rapid global increase in human
there were only about 100 such disasters
population should place more people in the
reported per year but that number has risen
path of natural events, raising both the
to over 300 a year since 2000. Experts say
number of events that are classified as
that increase is caused by climate change. In
natural disasters and their accompanying
contrast, natural geological disasters, such as
death tolls. In some areas, this is exactly
volcanic eruptions, earthquakes, and
what we are seeing.” So it's not only the
landslides, have remained steady in recent
climate that is changing, but also us humans.
decades.
MISLEADING HEADLINES
Drawn by cost pressure, undeveloped
land and fertile soil, people are flocking to
disaster-prone regions. People are tempting
nature with rapid urbanization in disasterprone regions, increasing the likelihood that
flash floods, storms, or earthquakes will
affect their towns and villages. This creates
a situation in which ordinary events like
earthquakes
and hurricanes
become
increasingly elevated to the level of natural
disasters that reap heavy losses in human
life and property.
In any given year, the death toll at the
hands of Mother Nature varies greatly, and
last year witnessed more fatalities than the
43
year before. However, over a longer time
span, the global number of deaths from
natural disasters has in fact decreased by 5%
in recent decades, thanks to better disaster
preparedness and prevention programs.
Hurricane Katrina in 2005 became the
costliest and one of the deadliest natural
disasters of all time, causing over $200
billion dollars in damage. In 2010, the
earthquake in Haiti, the heat wave in Russia,
and the floods in Pakistan were the biggest
killers, and deadly earthquakes also struck
Chile, Turkey, and China in one of the most
active seismic years in decades.
Opinion articles.
GM Foods may Pose Long-term Health Risks (incongruent) / GM foods are Safe
(congruent)
In 2050, the world population will have
Others argue that GM technology is
risen to about 9.2 billion compared to the 6.9
needed because it allows foods to grow and
billion in 2010. Therefore, global farm
survive in different climates. Due to the
output must increase substantially in order to
stronger breeds of crops being grown, the
feed the world in 2050. A crucial tool for
food quality of GM foods is improved, and
agriculture to address the food security issue
GM foods require less use of harmful
is GM (Genetically Modified) crops. To
pesticides because they are more resistant
meet that food demand we need to increase
against pests. In a recent statement issued by
our agricultural yields and increase the
a consortium of over a dozen international
efficiency of how plants take up nutrients. It
science academies, including the Australian
means growing plants that use less water to
Academy of Science, it was explained, "
produce the same output and improving
GM products have been used for many years
resistance to disease and pests.
and have been consumed without any
Irene Michaels from Organic Food
substantiated evidence of ill effects on
Science Australia says the GM argument
health. There is no logical reason they
cannot be justified, as all of the claims about
should be of any health concern, and their
higher yield and longer shelf life have not
safety has been confirmed by many peerbeen proven. “GM crops may have slightly
reviewed studies world-wide.”
increased levels of nutrients and vitamins
Crops with the highest percentage of
but in order to do that, scientists have to
genetic modifications worldwide are
interfere with the plants’ complex metabolic
soybeans, canola, corn, rice, sugar beet,
pathway. This technology can't deliver on its
potatoes and cotton. In Australia, the only
promises as it is still unknown whether the
GM crops produced are cotton and canola
genetically mutated food is safe to eat, as
with GM cotton making up 90% of the
long-term
health
impacts
remain
Cotton crop.
undetermined.”
Fears of Fluoride in Drinking Water (incongruent) / Fluoride Beneficial in Drinking Water
(congruent)
Fluoride is a mineral, which occurs
Others argue that adding fluoride to
naturally in minute quantities in most water
drinking water is an important element in
supplies, though at varying concentrations
promotion of dental health, especially in
throughout the world. Fluoride is also added
children. Two thirds of the Australian
to drinking water supplies to promote dental
population’s drinking water has been
health by preventing tooth damage and
fluoridated, which has led to a substantial
decay. It has been used for more than sixty
decrease in tooth decay over time in
years around the world, including over forty
comparison to non-fluoridated drinking
in Western Australia. While some European
areas. Many major health organizations
countries have discontinued its use in
including the Australian Dental Association
drinking water, the Perth Water Corporation
state that “The evidence supporting the
and WA health authorities are currently
safety, efficacy, and cost-effectiveness of
reviewing plans to extend the fluoridation
fluoridation of community water supplies
scheme to Perth’s semi-rural outer suburbs.
comes from multiple sources, covering 50
Many groups have objected to
years of legitimate peer-reviewed research
fluoridation over the years, with Robert Paul
including long term studies on large
from the Wanneroo Citizens’ Association
populations. Ongoing assessments are
claiming that it may be harmful and increase
stringently carried out, ensuring that the
the risk of bone cancer, osteoporosis, as well
concentration of fluoride does not exceed
as a host of other serious medical
1.5 ppm in drinking water systems. At this
complications. “It is hard to know whether
level, fluoride is absolutely safe.”
concentration levels of fluoride in the water
The population in Perth’s outer suburbs,
have exceeded the supposedly safe levels, as
in particular in the cities of Wanneroo and
it is tasteless and odourless in water.
Swan, and the Shire of SerpentineFluoride has been used as an insecticide and
Jarrahdale is rapidly growing. Many of these
rat poison. Putting it in drinking water poses
semi-rural areas are not connected to the
a health risk, and is an unethical type of
major metropolitan water supply, but rely on
mass medication that disregards a person’s
smaller, independent water supply networks.
freedom of choice.”
Rating Scales (all Articles)
1. How interesting did you find the preceding news article? (1: ‘very uninteresting’ to 5:
‘very interesting’)
2. How easy to read was the preceding news article? (1: ‘very difficult’ to 5: ‘very easy’)
3. How informative was the preceding news article? (1: ‘very uninformative’ to 5: ‘very
informative’)
4. Have you heard about the topic in the preceding news article before? (Yes/No)
5. How familiar are you with the topic discussed in the preceding news article? (1: ‘not
familiar at all’ to 5: ‘very familiar’)
6. [in manipulation-check study only] How accurately did the headline summarize the
preceding article? (1: ‘very misleading’ to 5: ‘entirely accurate’)
Recall Questions (all Articles)
1. Briefly summarize the article in a few sentences.
2. Summarize the main point of the article in no more than 10 words.
Recognition Questions
Burglary rate. Which of the following countries was NOT mentioned in the article as
having a lower burglary rate than Australia? (Japan / Italy / Canada / The Netherlands)
Natural disasters. Which natural disaster was mentioned specifically at the end of
the article? [Hurricane Katrina (2005) / Indian Ocean Tsunami (2004) / Black Saturday
Bushfire (2009)]
GM food. What are the only two GM crops produced in Australia? (Rice & Sugar
beet / Canola & Cotton / Corn & Soybeans / Wheat & Potato)
Fluoridation. Which of the following areas was NOT mentioned in the article as
rapidly growing in Perth’s outer suburbs? (The City of Wanneroo / The City of Rockingham /
The Shire of Serpentine-Jarrahdale / The City of Swan)
46
Inference Questions
Burglary rate.
1. After reading the burglary article, indicate in the percentage scale below how much
you think the level of burglaries will change over the next 3 years. (-25% to +25%)
2. What should happen to government funding for burglary prevention programs over
the next 3 years? (-25% to 25%)
3. If you owned an average home in an average suburb in Australia, what do you think
would be your lifetime risk of being burgled at least once? (0% to 30%)
Natural Disasters.
1. After reading the Natural Disaster article, indicate in the percentage scale below by
how much you think the death toll from Natural Disasters will change over the next
10 years. (-25% to 25%)
2. What should happen to government funding for disaster preparedness programs in
affected areas? (-25 to +25%)
3. Some budget life insurances exclude natural disasters as a cause of death. How much
more would you be willing to pay extra for the inclusion of natural disasters? (0% to
30%)
GM food.
1. After reading the GM foods article, indicate how much you think the consumption of
GM foods will add to Australia’s public health expenditure over the next 10 years. (0
to 30%)
2. What should happen to government funding for research into the health risks
associated with GM foods? (-25% to +25%)
3. How much would you pay extra at the grocery store to ensure you are buying nonGM foods? (0 to 30%)
Fluoridation.
1. After reading the Fluoride water article, indicate how much you think drinking
fluoridated water will add to the public health expenditure in Perth’s outer suburbs
over the next 10 years. (0 to 30%)
2. What should happen to government funding for research into the health risks
associated with water fluoridation? (-25% to +25%)
3. If you had the choice, how much extra would you be willing to pay for water that is
untreated and thus free of fluoride? (0 to 30%)
47
Appendix B—Example Stimuli used in Experiment 2 (to be placed in Online Supplement)
Condition: Person-in-photo: Bad / Headline: Congruent
Perth woman defrauds investors of $3 million
A Perth finance broker has been charged with defrauding investors of more than $3
million using a scheme that claimed to sell shares in a phoney land development called
‘Glendaleville’. Mrs Denise Carmel, 62, started the scheme in 2008, selling
investments in the alleged village-type development
for pensioners. While she paid “small” monthly
dividends to investors, she is thought to have gained
$3.26 million from the offences with which she has
been charged, with individual amounts ranging from
$40,000 to $400,000.
Several of the woman's victims, who ranged
between 53 and 73 years old, had as their only source
of income the age pension. Martha Clarke, a Hillarys
grandmother, was among the hardest hit, losing over
$300,000. Police have confirmed that the scheme was
organised “professionally and meticulously,” fooling
even experienced investors.
Mrs Carmel
Condition: Person-in-photo: Bad / Headline: Incongruent
Pensioners fleeced in $3 million fraud
A Hillarys grandmother was among the hardest hit in a recent investment fraud,
losing over $300,000. Mrs Martha Clarke had invested in a phoney land development
called ‘Glendaleville’. Victims of the scheme were mainly pensioners, between 53 and
73 years old, with investment amounts ranging from
$40,000 to $400,000. Police have confirmed that the
scheme was organised “professionally and
meticulously,” fooling even experienced investors.
A Perth finance broker has been charged with
defrauding investors. Mrs Denise Carmel, 62, started
the scheme in 2008, selling investments in the alleged
village-type development for pensioners. While she
paid “small” monthly dividends to investors, she is
thought to have gained $3.26 million from the
offences with which she has been charged.
Mrs Carmel
48
Condition: Person-in-photo: Good / Headline: Incongruent
Perth woman defrauds investors of $3 million
A Perth finance broker has been charged with defrauding investors of more than $3
million using a scheme that claimed to sell shares in a phoney land development called
‘Glendaleville’. Mrs Denise Carmel, 62, started the scheme in 2008, selling
investments in the alleged village-type development
for pensioners. While she paid “small” monthly
dividends to investors, she is thought to have gained
$3.26 million from the offences with which she has
been charged, with individual amounts ranging from
$40,000 to $400,000.
Several of the woman's victims, who ranged
between 53 and 73 years old, had as their only source
of income the age pension. Martha Clarke, a Hillarys
grandmother, was among the hardest hit, losing over
$300,000. Police have confirmed that the scheme was
organised “professionally and meticulously,” fooling
Mrs Clarke
even experienced investors.
Condition: Person-in-photo: Good / Headline: Congruent
Pensioners fleeced in $3 million fraud
A Hillarys grandmother was among the hardest hit in a recent investment fraud,
losing over $300,000. Mrs Martha Clarke had invested in a phoney land development
called ‘Glendaleville’. Victims of the scheme were mainly pensioners, between 53 and
73 years old, with investment amounts ranging from
$40,000 to $400,000. Police have confirmed that the
scheme was organised “professionally and
meticulously,” fooling even experienced investors.
A Perth finance broker has been charged with
defrauding investors. Mrs Denise Carmel, 62, started
the scheme in 2008, selling investments in the alleged
village-type development for pensioners. While she
paid “small” monthly dividends to investors, she is
thought to have gained $3.26 million from the
offences with which she has been charged.
Mrs Clarke
49
Tables
Table 1
Inference Score Intercorrelations in Experiment 1
Article
Intercorrelation
Fact 1
Fact 2
Opinion 1
Opinion 2
Items 1 and 2
.43*
.49*
.33*
.25^
Items 1 and 3
.24^
.25^
.45*
.32*
Items 2 and 3
.11
.04
.35*
.14
Note. * p < .05; ^ p < .10
Figure 1. Chart showing the structural similarity between Experiments 1 and 2. Congruent and incongruent headline conditions are presented in
separate box columns; time and associated processing steps go from top to bottom as indicated by the black arrows.
Figure 2. Appropriateness rating (1-5 scale) for headlines used in Experiment 1. Error bars
denote within-subject standard errors of the mean.
52
Figure 3. Memory performance (0-1 scale) in Experiment 1. Error bars denote within-subject
standard errors of the mean.
53
Figure 4. Inference scores (0-1 scale, where higher values indicate more reliance on the
article gist) in Experiment 1. Error bars denote within-subject standard errors of the mean.
54
Figure 5. Face-rating scores on valence traits (0-1 scale) in Experiment 2. Error bars indicate
within-subject standard errors of the mean.